|Publication number||US5926812 A|
|Application number||US 08/829,451|
|Publication date||Jul 20, 1999|
|Filing date||Mar 28, 1997|
|Priority date||Jun 20, 1996|
|Publication number||08829451, 829451, US 5926812 A, US 5926812A, US-A-5926812, US5926812 A, US5926812A|
|Inventors||Oliver A. Hilsenrath, Ron Carmel, Hagai Ariel|
|Original Assignee||Mantra Technologies, Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (4), Non-Patent Citations (4), Referenced by (218), Classifications (15), Legal Events (4)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application is based upon copending provisional patent application Ser. No. 60/018,800 filed Jun. 20, 1996 entitled "Personal Web Map System", which is incorporated herein by reference.
The field of the invention relates to computer implemented methods for determining the similarity of the contents of text documents. In particular, the field of the invention relates to applications of such methods to searching large databases, such as hypertext documents on the Internet, and to automatically producing personalized content without requiring a user or users to provide or update personalized searching criteria.
There are known techniques for the comparison of text files to determine similarity and difference of content, and for providing personalized content to a user. A common difficulty of present techniques is that they often return inappropriate matches or miss desired matches. For example, the list of documents retrieved by a typical Internet search engine usually contains a large fraction of irrelevant items and misses a large fraction of relevant items. Although sophisticated searching techniques employing artificial intelligence and the like can improve the quality of searching, these techniques usually require a lengthy training phase. Additionally, many present searching systems require the user to manually select personalized search criteria. This process is laborious, time consuming, and often requires the user to learn cryptic searching syntax. It also has the disadvantage that as the user's interests change with time, the personalized criteria must be manually modified.
In summary, conventional systems for searching large data bases do not contain simple and fast techniques for comparing documents and correctly determining their similarity. In particular, conventional database searching techniques lack the ability to permit documents to be compared quickly and to enable their similarity to be determined with high accuracy. Conventional methods also fail to provide simple database searching techniques that can identify documents of interest to a user or users without requiring the user or users to manually provide personalized searching criteria, and without requiring the user or users to update such criteria.
Accordingly, there is a need to provide a method for accurately and quickly comparing the contents of text documents and determining their similarity. It also would be desirable to provide such a method which may be used to implement an improved database searching system. What is also needed is a database searching system that automatically identifies documents of interest to a user or users without requiring the user or users to specify any search criteria. There is also a need for a system which provides automatic updates to the search criteria without requiring direct user intervention.
One aspect of the invention provides a computer-implemented method for comparing the textual data of two sets of documents. The method includes the step of extracting from the two sets of documents corresponding sets of document extract entries. In the preferred embodiment, each document extract entry comprises a weighted word histogram for its corresponding document, a list of people and a list of companies. The weighted word histogram comprises a set of histogram word records, each of which comprises a word from the document, a word score, a number of appearances of the word in the document, and a list of position indices for the word within the document.
The method further includes a step of generating from the sets of document extract entries corresponding sets of word clusters. Each word cluster comprises a cluster word list having N words, an N×N total distance matrix, and an N×N number of connections matrix. In the preferred embodiment, each number of connections matrix contains information about the connections between words in the document, and the total distance matrix contains information about the total distances between words in the document. The cluster word list is determined by recursively calling a procedure that returns a list of words within a predetermined distance from a given word in the document, and calculating the number of connections matrix by repeatedly calling a procedure that determines the number of connections in the document between words.
The preferred embodiment includes a step of grouping similar word clusters and combining the similar word clusters to form a single word cluster for each group. The grouping comprises evaluating a measure of cluster similarity between two word clusters, and placing them in a common group of similar word clusters if the measure of similarity exceeds a predetermined value. The step of evaluating cluster similarity comprises intersecting clusters to form subclusters and calculating a function of the subclusters. Similar word clusters are combined by concatenating their word lists to form a combined word list, merging their total distance matrices, and merging their word connection matrices.
The method of the invention further includes a step of determining a degree of similarity between clusters, in particular, between a cluster generated from the first set of word clusters and a cluster generated from the second set of word clusters. In the preferred embodiment, the determining step comprises intersecting two clusters from the two sets of clusters, thereby forming subclusters, and calculating a function of the subclusters. The function is a measure of similarity of the clusters and its calculation includes the determination of values such as a sum of diagonal matrix elements, a sum of off-diagonal matrix elements, and a sum of all matrix elements.
In the preferred embodiment, the method is implemented in a system to automatically identify database documents which are of interest to a given user or users. In this implementation, the method comprises the step of deriving the first set of documents from a local data storage device, such as a user's hard disk. This first set of documents is derived by selecting the set of files that are associated with the user or users. In this implementation the extracting step comprises converting the first set of documents to hypertext format. It is a unique feature of the present invention that the first set of documents is customized to a particular user and generated automatically without requiring the user to manually select search criteria. The method also comprises the step of deriving the second set of documents from a second data storage device. For example, the documents may be web documents retrieved via a network data connection. This application of the invention, therefore, provides fast and accurate searching to identify documents of interest to a particular user or users without any need for the user or users to specify what search criteria to use.
The foregoing and other features, aspects and advantages of the invention will become better understood with regard to the following description, appended claims and accompanying drawings in which:
FIG. 1 is a block diagram of the automatic database searching system of the present invention.
FIG. 2 is an implementation of the present invention in which the local documents and local module are contained on a user desktop machine, the network module is contained on a separate network machine, and the database documents are contained on Internet machines.
FIG. 3 is an implementation of the present invention in which the local documents, local module and network module are contained on a user desktop machine, and the database documents are contained on Internet machines.
FIG. 4 is an implementation of the present invention in which the local documents and local module are contained on a user desktop machine, and the network module and the database documents are contained on local area network machines.
FIG. 5 is an implementation of the present invention in which the local documents, local module, network module, and the database documents are all contained on a user desktop machine.
FIG. 6 is a schematic diagram outlining an aspect of the present invention in which extract entries are created from documents and word clusters are created from the extract entries.
FIG. 7 is a schematic representation of the data structure of an extract entry according to the present invention.
FIG. 8 is a schematic representation of the data structure of a weighted word histogram according to the present invention.
FIG. 9 is a schematic representation of the data structure of a word record according to the present invention.
FIG. 10 is a schematic representation of the data structure of a word cluster according to the present invention.
FIG. 11 is a schematic representation of the data structure of a cluster word list according to the present invention.
FIG. 12 is a schematic representation of the data structure of a total distance matrix according to the present invention.
FIG. 13 is a schematic representation of the data structure of a number of connections matrix according to the present invention.
FIG. 14 is a schematic diagram outlining an aspect of the present invention in which similar word clusters are combined into single word clusters.
FIG. 15 is a schematic diagram representing the manner in which cluster word lists are combined according to the present invention.
FIG. 16 is a schematic diagram representing the manner in which cluster matrices are combined according to the present invention.
Although the following detailed description contains many specifics for the purposes of illustration, anyone of ordinary skill in the art will appreciate that many variations and alterations to the following details are within the scope of the invention. Accordingly, the following preferred embodiment of the invention is set forth without any loss of generality to, and without imposing limitations upon, the claimed invention.
In a preferred embodiment of the invention, a method of efficiently and accurately comparing the contents of documents is implemented in an automatic database searching system, as shown in FIG. 1. The system may be decomposed into a local module 20 having access to local documents 22, and a network module 24 having access to database documents 26. In addition, local module 20 and network module 24 are in communication with each other. As will be described in detail in what follows, local module 20 extracts from local documents 22 a set of document extract files. In a post-processing step, the local module then generates word clusters from the document extract files. Local module 20 also performs additional post-processing to combine similar clusters. These clusters represent the content of the original documents. In a preferred embodiment, the local module operates automatically upon documents associated with a given user, thereby generating a personalized interest profile without requiring the user to manually specify a profile or update a profile. Local module 20 communicates the word cluster profile to network module 24. Network module 24 constructs search engine requests and retrieves selected documents from the database documents 26. These documents are then extracted and clusters are formed from the extracts just as in the local module. The interest profile clusters are then compared with the selected database document clusters to determine which of the selected database documents match the interest profile. The matching documents are then retrieved in full to the network machine for efficient browsing by the user at a later time. All the above is performed automatically and without any user intervention or configuration. In addition, no profile updating is needed since the user files will automatically change as the user's interests change.
The modules shown in FIG. 1 may be combined in various ways for different specific applications. For example, in an embodiment shown in FIG. 2, the local module and the local documents reside on a user desktop machine 28, the network module resides on a separate network machine 30, and the database documents reside on various database machines 32. FIG. 3 shows an embodiment wherein a user desktop machine 34 includes the network module in addition to the local module and local documents. The database documents reside on database machines 36, as in the previous embodiment.
FIG. 4 shows another embodiment wherein the database documents and network module both reside within a local area network (LAN) 40. A user desktop machine 38 contains the local module and local documents. In yet another embodiment, all four components, the local documents, the local module, the network module, and the database documents, reside on a single user desktop machine 42. The essential operation of the modules is largely independent of the specific embodiment chosen. Specific embodiments, however, also include certain implementation-specific operations, as will be seen in the descriptions of the particular embodiments.
The local module 20 uses local documents 22 to create document extract files and then to generate word clusters that represent the content of selected local documents. As shown schematically in FIG. 6, a set of documents 44 is used to create a set of document extract entries 46, which in turn is used to create a set of word clusters 48. In the preferred embodiment, each document gives rise to one extract entry. Each extract entry, however, can give rise to one or more word clusters, depending on the content of the extract entry and predefined parameters of the local module. For example, "Document 1" in FIG. 6 gives rise to "Extract Entry 1", which gives rise to "Word Cluster 1a" and "Word Cluster 1b". Note that FIG. 6 illustrates a case where the set of documents and the set of extract entries happen to have three elements. In general, of course, the sets may have any number of elements.
FIG. 7 shows the content of an extract entry. It contains document file information, a people list, a company list, and a weighted word histogram. Document file information includes such data as the file type, the directory path for the file, and the last time the file was extracted. The people list contains names of people found in the document, and the company list contains names of companies found in the document. The weighted word histogram contains a list of word records, as shown in FIG. 8 in the case where there happen to be seven word records. As shown in FIG. 9, each word record contains a word from the document, a word score which is a measure of the importance of the word within the document, the number of appearances of the word in the document, and a set of word position indices which contains the positions of the word in the document. The extraction method used to generate extract entries from documents will be described in detail later.
Based on the information contained within an extract entry, the local module generates one or more word clusters in a post-processing step. FIG. 10 shows the content of a word cluster. It contains a cluster word list, a total distance matrix, and a number of connections matrix. The cluster word list simply contains a list of words for the cluster, as shown in FIG. 11 in the case where the cluster happens to have six words. In general, of course, the cluster word list may have any number N of words.
The total distance matrix is shown in schematic form in FIG. 12. It contains N×N elements indexed by an ordered pair of integers (ij), where i and j range from 1 to N. The value of the (ij) element of the total distance matrix is the total distance between all appearances of word i and all appearances of the word j, when word i appears before word j. In this context, the distance between two words is taken to mean the number of words separating one from the other. For example, two adjacent words have a separation of one. The diagonal entries of the total difference matrix are used to store the scores for the words.
The number of connections matrix is shown in schematic form in FIG. 13. It contains N×N elements indexed by an ordered pair of integers (ij), where i and j range from 1 to N. The value of the (ij) element of the number of connections matrix is the total number of connections between all appearances of word i and all appearances of the word j, when word i appears before word j. In this context, a connection between two words is taken to mean an appearance of one word within a predetermined range of the other. A more specific description of the connection will be described below in relation to the method for calculating connections. The diagonal entries of the number of connections matrix are used to store the number of appearances of the words. The details of the method used to generate clusters from extract entries will be described in detail later.
In an additional post-processing step, the local module combines groups of similar clusters into single clusters, as shown schematically in FIG. 14 in the particular case where a set 50 of nine clusters are collected into three groups. The clusters in each group are then combined to form a set 52 with three clusters. Any two clusters may be combined by taking the union of their cluster word lists and merging their matrices.
For example, FIG. 15 is a schematic representation of how two cluster word lists are combined. In this example, a first cluster word list has four words, a second has five words, and the two lists have three words in common. Therefore, the new cluster word list contains six words: the three common words 54 corresponding to three words 56 in the first list and the three words 58 in the second list, one word 60 corresponding to the one remaining word 62 from the first cluster, and two words 64 corresponding to the two remaining words 66 from the second cluster.
FIG. 16 illustrates how the corresponding matrices from the same two clusters are merged. Since the clusters have three words in common, their respective matrices are divided into sub-matrices, as shown. The 4×4 matrix for the first cluster is divided into four sub-matrices: a 3×3 matrix 68 corresponding to the three words it shares with the second cluster, a 1×1 matrix 70 corresponding to the one word it does not share with the second cluster, a 1×3 matrix 72, and a 3×1 matrix 74. Similarly, the second cluster's 5×5 matrix is divided up into four sub-matrices: a 3×3 matrix 76 corresponding to the three words it shares with the first cluster, a 2×2 matrix 78 corresponding to the two words it does not share with the first cluster, a 3×2 matrix 80, and a 2×3 matrix 82. The first and second matrices are merged to form a 6×6 matrix, as shown. This 6×6 matrix contains a 3×3 matrix 84 whose entries contain the sum of the corresponding entries for 3×3 matrices 68 and 76. It also contains a 1×1 matrix 86 corresponding to the 1×1 matrix 70 from the first cluster, a 2×2 matrix 88 corresponding to the 2×2 matrix 78 from the second cluster, a 3×1 matrix 90 and a 1×3 matrix 92 corresponding to the 3×1 matrix 74 and the 1×3 matrix 72, and a 3×2 matrix 94 and a 2×3 matrix 96 corresponding to the 3×2 matrix 80 and the 2×3 matrix 82. The vacancies in the 6×6 matrix are set to zero.
Local Module--Automatic Operation
In the preferred embodiment, the local module is programmed to automatically and periodically examine local documents and generate a custom interest profile comprising word clusters. The local module is programmed to begin its operation automatically at times when the user is idle. These times may be determined by monitoring system use using various techniques such as recording the activation and deactivation times of a screen saver program. When the user is idle and is likely to remain idle based on past usage, the local module is activated. If the user returns to activity before the local module has completed its tasks, the local module operation is suspended until the computer becomes idle again. In this way, the operation of the local module does not consume computational resources when the computer is active.
Although the present system is unique in its capability to generate an interest profile without any direct manual input from or interaction with the user, this key feature does not preclude the user from manually and intentionally affecting the interest profile if desired. The present invention, therefore, may be easily adapted for use with present content provider systems to improve and augment their capabilities. This allowance for manual as well as automatic control of the profile is to be distinguished from prior art systems which depend upon and require manual input for their operation.
The local module is programmed to examine a local data storage device for files which are associated with a specified user. If the user is working within a LAN, then user files on many different machines can be included in the examination. The examined files may include the user's word processor documents, spreadsheets, electronic mail, preferences files, user interface files, and the like. If a file has not been modified since it was last extracted, it is not necessary to repeat the extraction process. Accordingly, those files which have been modified or created since the last extraction are marked for extraction. The examination updates information about all user files, such as the last time each file was read, the last time each was modified, the type of file, and so on.
The extraction process is applied to any file marked for extraction. Prior to the extraction proper, however, any file that is not already in hypertext markup language (HTML) format is converted to a temporary HTML file by one of many well-known standard HTML conversion methods. A document extraction procedure is then applied to each HTML document, thereby creating a new or updated document extract file entry for the document. The extract entry, as shown in FIG. 7, contains document file information, a people list, a company list, and a weighted word histogram. The structure of the weighted word histogram is described in detail above in relation to FIGS. 8 and 9.
The document extraction procedure identifies names within a document by searching for capitalized words after "Mr.", "Mrs.", "Ms.", "Prof.", "Dr.", etc., and for capitalized words before "Sr.", "Jr.", "Ph.D.", "J.D.", "M.D.", "CEO", etc. Similarly, it identifies company names by searching for capitalized words followed by "Co.", "Inc.", "Ltd.", "Corp.", "Technologies" etc. In generating the weighted word histogram, the procedure either ignores or deletes from the document all words such as "a", "the", "and", "it", "some", "him", "her", "it", etc. The weighted word histogram contains a word record for each distinct word present in the document. Each word record contains the word or a code for the word. Each record also contains a word score which is a measure of the importance of the word within the document. The word score is determined from the hypertext tags associated with the word. For example, if a word appears in a heading, it is given a higher score than if the word appears in a regular paragraph. Various HTML tags are assigned predetermined values. The word record also contains a number of appearances of the word in the document, and its position indices. The position indices for a word are simply the word positions of each appearance of the word in the document. Note that it is possible to include other significant information about words in the document, or to combine various forms of information together to create composite weights. For example, the word scores for a document can be decreased if the document has not been accessed in a long time, increased by a small amount if the document has been recently read, and increased by a large amount if the document has been recently modified.
When the extraction procedure is complete, the extract entries for all the user's files have been updated to reflect any changes since the last extraction. All temporary HTML files and other files related to the documents are not generally needed for subsequent steps. Only the extract entries are used.
Local Module--Generation of Word Clusters
In a post-processing stage executed after the document extract file has been updated, the local module creates one or more document word clusters from each extract entry. Word clusters contain information about the importance of various words in the document and their relationships to each other within the document, as will become clear from the following description. The post-processing stage further includes combining similar clusters to form composite clusters.
As described above in relation to FIGS. 10-13, a word cluster comprises a cluster word list, a total distance matrix, and a number of connections matrix. The cluster word list is a list of N words in the cluster. The total distance matrix is an N×N matrix whose (ij) element contains a total distance between word i and word j when word i precedes word j. The diagonal element (i,i) contains the score of word i in the document. The number of connections matrix is an N×N matrix whose (ij) element contains a number of connections between word i and word j when word i precedes word j. The diagonal element (i,i) contains the number of appearances of word i in the document.
In the post-processing stage, a set of one or more initial word clusters is produced from each document extract entry, then similar clusters are merged. An initial cluster is formed for the most significant words in the weighted word histogram for the document. The significance of a word is derived from the score and/or number of appearances of the word. Other measures of significance may obviously be used without departing from the scope of the invention.
An initial word cluster for a word is constructed as follows. First, a cluster word list is determined from document extract entry by recursively calling a procedure that returns a list of other words within a predetermined distance from the word. For example, if the initial word is "mountain", then a procedure will examine the weighted word histogram and return a list of other words within a predefined number of words to the right and left of each appearance of "mountain". This predefined number of words is called the breadth. Suppose it found two such words: "bike" and "trail". The procedure is then called again for each of these two words and returns a list of other words close to "bike", and a list of other words close to "trail". The cluster word list is simply a list of the words found through this recursive procedure, in this example, "mountain", "bike", "trail", and the others. In the preferred embodiment, the recursive depth is limited to a predefined number called the depth. It is also terminated in the case where no new word is found. There are thus two parameters which are used in determining a cluster word list: the breadth of the search and the depth of the search. The breadth is a measure of how closely positioned words must be, and the depth is a measure of the maximum number of degrees of separation (i.e. recursive calls) between the words. If a first and second word are within breadth words of each other in at least one of their appearances, it is said that they are connected.
If there are N words in the cluster word list, the number of connections matrix contains N×N elements. The (ij) element contains the total number of connections between word i and word j when word i appears before word j. For example, if "mountain bike" appears ten times in a document, then the number of connections of "mountain" to "bike" is at least ten. Note that this number is usually different from the number of connections of "bike" to "mountain" since the order of the words is taken into account. Accordingly, two words are considered to be closely associated with each other if their number of connections in one order differs from their number of connections in another order. The total number of connections in either order, of course, also indicates that they are closely associated.
The total distance matrix also has N×N elements. The (ij) element contains the sum of the distances between word i and word j for all connections of word i with word j when word i appears before word j. Thus, the ratio of total distance matrix element (ij) with number of connections matrix element (ij) is the average distance between word i and word j for all connections between word i and word j when word i appears before word j. The total distance matrix, as well as the number of connections matrix, are easily calculated based on the above descriptions and techniques well known in the art.
Local Module--Grouping and Combination of Word Clusters
The initial word clusters that are generated from a document extract entry typically fall into various natural groups which roughly correspond to topics or subject areas of interest to the user. The purpose of the next stage in the post-processing is to compare the initial word clusters with each other to determine their degrees of similarity, to group similar clusters together, and to merge the similar clusters into a single cluster for each group.
The comparison of one cluster with another is performed as follows. First, an intersection of the clusters is determined by finding the words shared in common by their cluster word lists. For example, FIG. 15 shows two clusters that share three words in common. These common words partition the cluster word lists into shared and non-shared words. They also partition each cluster matrix into four submatrices, a square matrix corresponding to the shared words, a square matrix corresponding to the non-shared words, and two remaining hybrid matrices. An example of this division of the matrices is shown in FIG. 16. Note that a rearrangement of the rows and columns of the matrices may be necessary for them to appear as shown in FIG. 16. Such a rearrangement, however, is not actually required.
After the matrices of the two clusters have been divided as described above, their degree of similarity can be determined as follows. A measure of similarity is calculated using certain values derived from the eight submatrices from the two clusters. Examples of such values are (1) the sum of the diagonal elements in one of the four square matrices, (2) the sum of the off-diagonal elements in one of the four square matrices, and (3) the sum of all elements in one of the four hybrid matrices. These various quantities give information about the degree of similarity between the two clusters and can be combined in various ways to construct measures of cluster similarity. For example, let C1 be the sum of the diagonal elements of the shared matrix 68 for the first cluster. C1 is then a measure of the total number of appearances of the shared words in the first document. Let D1 be the sum of the diagonal elements of the non-shared matrix 70 for the first cluster. D1 is then the total number of appearances of the non-shared words in the first document. For closely related clusters, C1 should be large in comparison with D1. The same should be true of the corresponding values of C2 and D2 for the second cluster. Thus, three possible measures of similarity of the two clusters are
S=(C1 +C2)/(D1 +D2),
S=C1 /D1 +C2 /D2,
S"=0.5 C1 /(C1 +D1)+C2 /(C2 +D2)!.
The advantage of the last measure, S" is that it has been constructed so that it always takes values between 0 and 1, where a value of 1 corresponds to identical clusters and a value of 0 corresponds to entirely distinct clusters. For more subtle and refined measures of similarity, the off-diagonal elements and the hybrid matrices can also be included into the calculation. Other measures can use the total distance matrix, the number of connections matrix, or both in combination. It is an important aspect of the present invention that the clusters, and these matrices in particular, allow for the construction and use of very subtle measures of similarity between documents.
If the measure of similarity of two clusters exceeds a predetermined threshold value, then the two clusters are grouped together. Two clusters in the same group are merged to form a single cluster as illustrated in FIG. 16. The new cluster word list comprises the union of the word lists of the two original clusters, and the matrices of the two clusters are merged as shown in the figure and as described in detail previously.
Local Module--Transfer Clusters to Network Module
After all the similar word clusters have been merged, the resulting clusters represent distinct "subject areas of interest" for the user whose files were extracted. The local module then transfers these word clusters to the network module. The local module has then completed its cycle and it now awaits activation for the next automatic update.
In embodiments where the local module communicates with the network module via a network connection (e.g. the embodiment shown in FIG. 2), it is useful to have a method for transferring the cluster data from the client machine to the network machine using standard communication tools. Accordingly, an aspect of the present invention includes the following novel and advantageous technique for transmitting the cluster data to the network machine. The technique includes imbedding cluster data into fields contained in an HTML form submission. In conventional HTML form submission, a client requests a form from the server, the user manually updates values in the page, and returns it to the server. The present technique, in contrast, automatically embeds the cluster data into a form and submits it to the network machine without any user intervention. The entire process is transparent to the user. In a preferred embodiment of the invention, the cluster data may be compressed, encrypted, or both.
The network module has the basic task of determining matches between the database documents and the cluster data it has received from the local module. In the preferred embodiment, the database documents are stored on a network. The network module uses the cluster data to generate search strings which are then submitted to one or more third party search engines. The network module retrieves the documents corresponding to the best matches and then generates corresponding document extract entries. The network module then generates a set of word clusters for each extract entry. These clusters are then compared with the clusters from the local module using a measure of similarity. If a database document cluster is sufficiently similar to a cluster from the local module, then the database document is considered to be a match with the local documents. The document is then downloaded to the network machine together with its associated resource files. The result is a collection of documents located at the network machine. These documents are thus available for fast and convenient access by the local module. This predictive caching technique is unique in that the documents are retrieved by and stored at the network machine before the user requests a document for the first time. As a result, the benefits of caching are obtained even on the first access by the user. Conventional non-predictive caching, in contrast, has benefits only after the first access. The preferred embodiment also includes the convenient feature of organizing the personalized content for the user, e.g. in the form of a table of contents organized by subject.
Network Module--Generating Search Strings
After the network module receives a set of new word clusters from the local module its first step is to generate search engine submission strings based on the information in the clusters. The technique for generating these strings and retrieving documents using existing search engines is an important aspect of the present invention. This technique is able to avoid the inaccuracies in the existing search engines through the use of supplementary and more exact filtering which involves document extraction, cluster generation, and cluster similarity measurement.
For each cluster received from the local module, the network module creates a series of search strings of varying degrees of scope, as follows. A number M of most important words in the cluster is found by selecting those words from the cluster word list which have traits reflecting higher significance, e.g. a score larger than a predetermined threshold value. Alternatively, they could be chosen by how many appearances they have in the document, or by some combination of score and number of appearances that correlates with importance. These M important words, w1, w2, . . . , wM-1, wM, can be combined to produce search strings varying from narrow to broad, as follows:
S1 : w1 and w2 and w3 and . . . and wM-2 and wM-1 and wM
S2 : w1 and w2 and w3 and . . . and wM-2 and (wM-1 or wM)
S3 : w1 and w2 and w3 and . . . and (wM-2 or wM-1 or wM)
SM-2 : w1 and w2 and (w3 or . . . or wM-2 or wM-1 or wM)
SM-1 : w1 and (w2 or w3 or . . . or wM-2 or wM-1 or wM)
SM : (w1 or w2 or w3 or . . . or wM-2 or wM-1 or wM)
Other search strings can also be created using connectives such as "near". These connectives could be used, if necessary, to form strings of more or less restrictive scope.
The search strings are submitted to the search engine as follows. First, the most narrow string, S1, is submitted to the engine and the document references, e.g. uniform resource locators (URLs), for the matches are added to a list of preliminary search results. If the number of document references in the list is less than a predetermined number of desired document references, then the next search string, S2, which is slightly broader in scope than S1, is submitted to the search engine. Any new document references that have not already been found are added to the list. If the number of document references in the list is still less than the predetermined number, then the next string, S3, is submitted and the new document references are added to the list. This process continues until the desired number or document references is obtained.
Network Module--Retrieval, Extraction, and Cluster Generation
Once a sufficiently large list of document references has been obtained from the search engine, the network module attempts to retrieve the documents corresponding to these document references. Since some document references may be outdated, not all the documents can be retrieved. Consequently, the list of retrievable document references is a subset of the original list. If the retrieved documents are not already hypertext, they are converted to HTML by a conventional HTML converter. The network module then generates document extracts and clusters in the same manner as the local module, just as described above in relation to FIGS. 6-13. The result is a collection of word clusters associated with each of the database documents in the list of documents retrieved.
It is important to note that the local module transmits to the network module information indicating which clusters are new or modified and which are unchanged. Accordingly, the network module can intelligently eliminate unnecessary repetition of processing by taking into consideration the processing which has already taken place. For example, if only one cluster in the interest profile has been changed, then it may be advisable to only perform searches based on words from this modified cluster. Due to changes in the content of the database, however, periodic searches based on all the interest clusters may be performed, regardless of whether or not the interest profile has been changed.
Network Module--Document Relevance
After the network module has obtained clusters for the database documents retrieved, it then eliminates irrelevant documents by comparing the database clusters with the interest profile clusters and retaining only those documents whose clusters are sufficiently similar to one of the profile clusters. The network module determines similarity of clusters in the same way as the local module, just as described in detail above. A database document is discarded as irrelevant if, when its clusters are compared with the interest profile clusters, the similarity measure is never above a predefined minimum threshold.
In another embodiment, the search engine is adapted for searching based directly on cluster data rather than search strings. In this case, the generation of search strings from the clusters is by-passed entirely, and relevant documents are directly determined and returned from the search engine.
Network Module--Caching and Creating Custom Interface
In the next stage of its operation, the network module retrieves all the relevant database documents for the user together with their associated resource files, and creates a customized user interface which organizes the documents for convenient browsing. The result is a collection of documents at the network machine which are customized to the user and available for fast and convenient access. Because the documents have been transplanted to the network machine, however, certain internal document references must be changed so that, for example, a document properly refers to its resource files at the network machine rather than at the native web site of the document. Techniques for modifying internal document references (e.g. URLs) for this purpose are well-known in the art.
The user interface organizes the personalized content for the user in a form that is specially adapted to the user's preferences. For example, the content may be presented in the form of a table of contents organized by subject, or as a collection of icons, or a combination of words and icons. The form in which the information is organized and presented can be manually selected by the user, or can be automatically determined by the system based upon preferences information in the local documents. This can include such parameters as whether the user prefers to view information by icons or by words. The interface can also be automatically adjusted in the amount of detail it includes, depending on whether the user has more or less time to explore the information.
Obviously, the network machine can handle many different users at once, and is not limited to serving just one user or local module. In the case where a collection of users is known to have a large area of overlapping interests, the network module can save valuable storage space, processing time and network bandwidth on the network machine by performing collective caching for the group of users. For example, an academic department at a university may have its own network machine which performs collective caching for its users who are researchers in the same academic discipline.
In another embodiment, the database documents are documents existing on one or more local storage devices. For example, while a user is writing a research paper, article or book, the system automatically constructs a citation list that provides references between the research paper and other documents stored locally (i.e. on the desktop machine or on a LAN connected to the desktop machine). These local documents may include articles, E-mail, reference materials, and other documents. The system thus actively and automatically alerts the user of local information related to present work.
While the invention has been described in connection with what is presently considered to be the most practical and preferred embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but on the contrary it is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4839853 *||Sep 15, 1988||Jun 13, 1989||Bell Communications Research, Inc.||Computer information retrieval using latent semantic structure|
|US5619709 *||Nov 21, 1995||Apr 8, 1997||Hnc, Inc.||System and method of context vector generation and retrieval|
|US5724571 *||Jul 7, 1995||Mar 3, 1998||Sun Microsystems, Inc.||Method and apparatus for generating query responses in a computer-based document retrieval system|
|US5754938 *||Oct 31, 1995||May 19, 1998||Herz; Frederick S. M.||Pseudonymous server for system for customized electronic identification of desirable objects|
|1||"Assignment of Term Descriptors to Clusters,", Bhatia et al., Proceddings of the 1990 Symposium on Applied Computing, USA, IEEE Computer Society, pp. 181-185, Apr. 1990.|
|2||"Search and Ranking Algorithms for Locating Resources on the World Wide Web," Yuwono et al., Procedings of the Twelfth International Conference on Data Engineering, CA, USA, pp. 164-171, IEEE, Feb. 1996.|
|3||*||Assignment of Term Descriptors to Clusters, , Bhatia et al., Proceddings of the 1990 Symposium on Applied Computing, USA, IEEE Computer Society, pp. 181 185, Apr. 1990.|
|4||*||Search and Ranking Algorithms for Locating Resources on the World Wide Web, Yuwono et al., Procedings of the Twelfth International Conference on Data Engineering, CA, USA, pp. 164 171, IEEE, Feb. 1996.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US6144973 *||Sep 5, 1997||Nov 7, 2000||Kabushiki Kaisha Toshiba||Document requesting system and method of receiving related document in advance|
|US6167397 *||Sep 23, 1997||Dec 26, 2000||At&T Corporation||Method of clustering electronic documents in response to a search query|
|US6292796 *||Feb 23, 1999||Sep 18, 2001||Clinical Focus, Inc.||Method and apparatus for improving access to literature|
|US6338060 *||Jan 26, 1999||Jan 8, 2002||Canon Kabushiki Kaisha||Data processing apparatus and method for outputting data on the basis of similarity|
|US6347315 *||Dec 8, 1998||Feb 12, 2002||Canon Kabushiki Kaisha||Method and apparatus for selecting and utilizing one of computers or databases|
|US6363379||Sep 28, 2000||Mar 26, 2002||At&T Corp.||Method of clustering electronic documents in response to a search query|
|US6430559 *||Nov 2, 1999||Aug 6, 2002||Claritech Corporation||Method and apparatus for profile score threshold setting and updating|
|US6463434 *||Dec 17, 2001||Oct 8, 2002||Claritech Corporation||Method and apparatus for profile score threshold setting and updating|
|US6493711 *||May 5, 1999||Dec 10, 2002||H5 Technologies, Inc.||Wide-spectrum information search engine|
|US6505168||Aug 16, 1999||Jan 7, 2003||First Usa Bank, Na||System and method for gathering and standardizing customer purchase information for target marketing|
|US6505196 *||Aug 14, 2001||Jan 7, 2003||Clinical Focus, Inc.||Method and apparatus for improving access to literature|
|US6535876 *||Jun 5, 2002||Mar 18, 2003||Clairvoyance||Method and apparatus for profile score threshold setting and updating|
|US6587850 *||Jun 5, 2002||Jul 1, 2003||Claritech Corporation||Method and apparatus for profile score threshold setting and updating|
|US6629097||Apr 14, 2000||Sep 30, 2003||Douglas K. Keith||Displaying implicit associations among items in loosely-structured data sets|
|US6633868||Jul 28, 2000||Oct 14, 2003||Shermann Loyall Min||System and method for context-based document retrieval|
|US6654735 *||Jan 8, 1999||Nov 25, 2003||International Business Machines Corporation||Outbound information analysis for generating user interest profiles and improving user productivity|
|US6654739 *||Jan 31, 2000||Nov 25, 2003||International Business Machines Corporation||Lightweight document clustering|
|US6708165 *||Dec 10, 2002||Mar 16, 2004||H5 Technologies, Inc.||Wide-spectrum information search engine|
|US6711568||Nov 8, 2000||Mar 23, 2004||Krishna Asur Bharat||Method for estimating coverage of web search engines|
|US6711586||Jul 17, 2000||Mar 23, 2004||William Mitchell Wells||Methods and systems for providing information based on similarity|
|US6741983 *||Sep 28, 2000||May 25, 2004||John D. Birdwell||Method of indexed storage and retrieval of multidimensional information|
|US6785684 *||Mar 27, 2001||Aug 31, 2004||International Business Machines Corporation||Apparatus and method for determining clustering factor in a database using block level sampling|
|US6826574 *||Aug 27, 1999||Nov 30, 2004||Gateway, Inc.||Automatic profiler|
|US6829615||Feb 25, 2000||Dec 7, 2004||International Business Machines Corporation||Object type relationship graphical user interface|
|US6931399||Jun 26, 2001||Aug 16, 2005||Igougo Inc.||Method and apparatus for providing personalized relevant information|
|US6947933 *||Dec 17, 2003||Sep 20, 2005||Verdasys, Inc.||Identifying similarities within large collections of unstructured data|
|US7010515||Jul 12, 2002||Mar 7, 2006||Matsushita Electric Industrial Co., Ltd.||Text comparison apparatus|
|US7062487 *||Jun 2, 2000||Jun 13, 2006||Seiko Epson Corporation||Information categorizing method and apparatus, and a program for implementing the method|
|US7072902 *||Nov 26, 2002||Jul 4, 2006||Tzunami Inc||Method and system for organizing objects according to information categories|
|US7089194 *||Jun 17, 1999||Aug 8, 2006||International Business Machines Corporation||Method and apparatus for providing reduced cost online service and adaptive targeting of advertisements|
|US7113943||Dec 5, 2001||Sep 26, 2006||Content Analyst Company, Llc||Method for document comparison and selection|
|US7171620 *||Jul 24, 2002||Jan 30, 2007||Xerox Corporation||System and method for managing document retention of shared documents|
|US7260773 *||Mar 17, 2003||Aug 21, 2007||Uri Zernik||Device system and method for determining document similarities and differences|
|US7272612||Jan 30, 2004||Sep 18, 2007||University Of Tennessee Research Foundation||Method of partitioning data records|
|US7308445 *||Jan 21, 2004||Dec 11, 2007||Overture Services, Inc.||Method for estimating coverage of web search engines|
|US7386545||Mar 31, 2005||Jun 10, 2008||International Business Machines Corporation||System and method for disambiguating entities in a web page search|
|US7433883||Mar 12, 2004||Oct 7, 2008||H5 Technologies||Wide-spectrum information search engine|
|US7444325 *||Apr 25, 2005||Oct 28, 2008||Im2, Inc.||Method and system for information extraction|
|US7454411||Jan 30, 2004||Nov 18, 2008||Universtiy Of Tennessee Research Foundation||Parallel data processing architecture|
|US7490116||Dec 17, 2003||Feb 10, 2009||Verdasys, Inc.||Identifying history of modification within large collections of unstructured data|
|US7536433||Sep 25, 2003||May 19, 2009||Jpmorgan Chase Bank, N.A.||System and method for customizing a portal environment|
|US7624087||Nov 16, 2006||Nov 24, 2009||University Of Tennessee Research Foundation||Method of expert system analysis of DNA electrophoresis data|
|US7640223||Nov 16, 2006||Dec 29, 2009||University Of Tennessee Research Foundation||Method of organizing and presenting data in a table using stutter peak rule|
|US7661064 *||Mar 6, 2006||Feb 9, 2010||Microsoft Corporation||Displaying text intraline diffing output|
|US7664719||Nov 16, 2006||Feb 16, 2010||University Of Tennessee Research Foundation||Interaction method with an expert system that utilizes stutter peak rule|
|US7685013||Aug 24, 2007||Mar 23, 2010||Jpmorgan Chase Bank||System and method for automatic financial project management|
|US7689504||Oct 31, 2002||Mar 30, 2010||Jpmorgan Chase Bank, N.A.||System and method for establishing or modifying an account with user selectable terms|
|US7689506||Jun 7, 2002||Mar 30, 2010||Jpmorgan Chase Bank, N.A.||System and method for rapid updating of credit information|
|US7702680 *||Nov 2, 2006||Apr 20, 2010||Microsoft Corporation||Document summarization by maximizing informative content words|
|US7711738 *||Nov 15, 2000||May 4, 2010||West Services, Inc.||Method, system and computer-readable medium for accessing and retrieving court records, items and documents|
|US7734627 *||Jun 17, 2003||Jun 8, 2010||Google Inc.||Document similarity detection|
|US7742997||Apr 23, 2004||Jun 22, 2010||Jpmorgan Chase Bank, N.A.||System and method for management and delivery of content and rules|
|US7747866||Aug 24, 2000||Jun 29, 2010||Jpmorgan Chase Bank, N.A.||System and method providing improved error detection related to transmission of data over a communication link|
|US7756816||Oct 12, 2005||Jul 13, 2010||Jpmorgan Chase Bank, N.A.||System and method for network-based project management|
|US7756855||Aug 28, 2007||Jul 13, 2010||Collarity, Inc.||Search phrase refinement by search term replacement|
|US7769803||Jan 2, 2008||Aug 3, 2010||University Of Tennessee Research Foundation||Parallel data processing architecture|
|US7783578||Jan 25, 2006||Aug 24, 2010||Jpmorgan Chase Bank, N.A.||System for providing cardless payment|
|US7831508||Jun 23, 2000||Nov 9, 2010||Jpmorgan Chase Bank, N.A.||System and method for implementing a consolidated application process|
|US7840519||Nov 13, 2009||Nov 23, 2010||University Of Tennesse Research Foundation||Organizing and outputting results of a DNA analysis based on firing rules of a rule base|
|US7844141 *||Dec 1, 2008||Nov 30, 2010||Gary Stephen Shuster||Method and apparatus for photograph finding|
|US7844566||May 11, 2006||Nov 30, 2010||Content Analyst Company, Llc||Latent semantic clustering|
|US7882106||Aug 10, 2007||Feb 1, 2011||University Of Tennessee Research Foundation||Method of indexed storage and retrieval of multidimensional information|
|US7929810||Sep 2, 2010||Apr 19, 2011||Hoshiko Llc||Method and apparatus for photograph finding|
|US7941533||Feb 19, 2002||May 10, 2011||Jpmorgan Chase Bank, N.A.||System and method for single sign-on session management without central server|
|US7962486 *||Jan 10, 2008||Jun 14, 2011||International Business Machines Corporation||Method and system for discovery and modification of data cluster and synonyms|
|US7966496||Dec 22, 2006||Jun 21, 2011||Jpmorgan Chase Bank, N.A.||System and method for single sign on process for websites with multiple applications and services|
|US7987501||Dec 21, 2001||Jul 26, 2011||Jpmorgan Chase Bank, N.A.||System and method for single session sign-on|
|US8060522||Aug 27, 2008||Nov 15, 2011||University Of Tennessee Research Foundation||Parallel data processing system|
|US8099733||Jan 2, 2008||Jan 17, 2012||Birdwell John D||Parallel data processing architecture|
|US8122032||Jul 20, 2007||Feb 21, 2012||Google Inc.||Identifying and linking similar passages in a digital text corpus|
|US8145522||Nov 8, 2006||Mar 27, 2012||Jpmorgan Chase Bank, N.A.||System and method for establishing or modifying an account with user selectable terms|
|US8151186 *||Sep 8, 2011||Apr 3, 2012||Google Inc.||Comparing text pages using image features based on word positions|
|US8151187 *||Sep 26, 2011||Apr 3, 2012||Google Inc.||Comparing text pages using image features based on word positions|
|US8160960||Dec 11, 2009||Apr 17, 2012||Jpmorgan Chase Bank, N.A.||System and method for rapid updating of credit information|
|US8185877||Jun 22, 2005||May 22, 2012||Jpmorgan Chase Bank, N.A.||System and method for testing applications|
|US8185940||Jul 17, 2007||May 22, 2012||Jpmorgan Chase Bank, N.A.||System and method for providing discriminated content to network users|
|US8190893||Jul 1, 2004||May 29, 2012||Jp Morgan Chase Bank||Portable security transaction protocol|
|US8209339||Apr 21, 2010||Jun 26, 2012||Google Inc.||Document similarity detection|
|US8239415||Mar 8, 2011||Aug 7, 2012||Lexisnexis, A Division Of Reed Elsevier Inc.||Systems and methods for providing a validation tool|
|US8245193||May 13, 2008||Aug 14, 2012||Chipmos Technologies Inc||Automatic creative proposal generating and filtering system and manufacturing method thereof and multiple components combining method|
|US8266054||Jan 19, 2012||Sep 11, 2012||Jpmorgan Chase Bank, N.A.||System and method for implementing a consolidated application process|
|US8266437||Jun 25, 2002||Sep 11, 2012||Jpmorgan Chase Bank, N.A.||Electronic vouchers and a system and method for issuing the same|
|US8301493||Nov 5, 2002||Oct 30, 2012||Jpmorgan Chase Bank, N.A.||System and method for providing incentives to consumers to share information|
|US8306907||May 30, 2003||Nov 6, 2012||Jpmorgan Chase Bank N.A.||System and method for offering risk-based interest rates in a credit instrument|
|US8306987||Mar 4, 2009||Nov 6, 2012||Ofer Ber||System and method for matching search requests and relevant data|
|US8315483||Feb 28, 2012||Nov 20, 2012||Hoshiko Llc||Method and apparatus for photograph finding|
|US8321197||Oct 17, 2007||Nov 27, 2012||Teresa Ruth Gaudet||Method and process for performing category-based analysis, evaluation, and prescriptive practice creation upon stenographically written and voice-written text files|
|US8321682||Jan 26, 2009||Nov 27, 2012||Jpmorgan Chase Bank, N.A.||System and method for generating and managing administrator passwords|
|US8335855||May 18, 2007||Dec 18, 2012||Jpmorgan Chase Bank, N.A.||System and method for portal infrastructure tracking|
|US8385691||Apr 19, 2011||Feb 26, 2013||Hoshiko Llc||Method and apparatus for photograph finding|
|US8429184 *||Jun 14, 2010||Apr 23, 2013||Collarity Inc.||Generation of refinement terms for search queries|
|US8438086||Nov 28, 2001||May 7, 2013||Jpmorgan Chase Bank, N.A.||System and method for providing customers with seamless entry to a remote server|
|US8438178||Jun 25, 2009||May 7, 2013||Collarity Inc.||Interactions among online digital identities|
|US8442972||Oct 11, 2007||May 14, 2013||Collarity, Inc.||Negative associations for search results ranking and refinement|
|US8458070||Jun 15, 2012||Jun 4, 2013||Jpmorgan Chase Bank, N.A.||System and method for providing customers with seamless entry to a remote server|
|US8473735||May 19, 2008||Jun 25, 2013||Jpmorgan Chase||Systems and methods for managing digital certificates|
|US8549315||Sep 14, 2012||Oct 1, 2013||Jpmorgan Chase Bank, N.A.||System and method for generating and managing administrator passwords|
|US8571975||Nov 24, 1999||Oct 29, 2013||Jpmorgan Chase Bank, N.A.||System and method for sending money via E-mail over the internet|
|US8572089 *||Dec 15, 2011||Oct 29, 2013||Business Objects Software Ltd.||Entity clustering via data services|
|US8577864||Aug 2, 2012||Nov 5, 2013||Lexisnexis, A Division Of Reed Elsevier, Inc.||Systems and methods for providing a validation tool|
|US8583926||Apr 26, 2006||Nov 12, 2013||Jpmorgan Chase Bank, N.A.||System and method for anti-phishing authentication|
|US8590008||Oct 30, 2008||Nov 19, 2013||Jpmorgan Chase Bank, N.A.||System and method for single sign on process for websites with multiple applications and services|
|US8650199||Jun 25, 2012||Feb 11, 2014||Google Inc.||Document similarity detection|
|US8676746 *||Jan 3, 2008||Mar 18, 2014||Microsoft Corporation||Database management system risk assessment|
|US8712932||Jul 6, 2012||Apr 29, 2014||Chipmos Technologies Inc.||Computer implemented apparatus for generating and filtering creative proposal|
|US8726011||May 17, 2013||May 13, 2014||Jpmorgan Chase Bank, N.A.||Systems and methods for managing digital certificates|
|US8732072||Mar 16, 2010||May 20, 2014||Jpmorgan Chase Bank, N.A.||System and method for establishing or modifying an account with user selectable terms|
|US8782067 *||Mar 23, 2012||Jul 15, 2014||Casio Computer Co., Ltd||Searching method, searching device and recording medium recording a computer program|
|US8793490||Dec 13, 2006||Jul 29, 2014||Jpmorgan Chase Bank, N.A.||Systems and methods for multifactor authentication|
|US8805739||Mar 23, 2001||Aug 12, 2014||Jpmorgan Chase Bank, National Association||System and method for electronic bill pay and presentment|
|US8812541||Mar 12, 2013||Aug 19, 2014||Collarity, Inc.||Generation of refinement terms for search queries|
|US8831969 *||Oct 2, 2013||Sep 9, 2014||Linkedin Corporation||System and method for determining users working for the same employers in a social network|
|US8832057 *||Dec 2, 2011||Sep 9, 2014||Yahoo! Inc.||Results returned for list-seeking queries|
|US8832102 *||Jan 12, 2010||Sep 9, 2014||Yahoo! Inc.||Methods and apparatuses for clustering electronic documents based on structural features and static content features|
|US8849716||Sep 14, 2007||Sep 30, 2014||Jpmorgan Chase Bank, N.A.||System and method for preventing identity theft or misuse by restricting access|
|US8867798||Feb 25, 2013||Oct 21, 2014||Intellectual Ventures I Llc||Method and apparatus for photograph finding|
|US8875038||Jan 19, 2011||Oct 28, 2014||Collarity, Inc.||Anchoring for content synchronization|
|US8892574 *||Nov 6, 2009||Nov 18, 2014||Nec Corporation||Search apparatus, search method, and non-transitory computer readable medium storing program that input a query representing a subset of a document set stored to a document database and output a keyword that often appears in the subset|
|US8903810||Oct 16, 2008||Dec 2, 2014||Collarity, Inc.||Techniques for ranking search results|
|US8909627||Oct 26, 2012||Dec 9, 2014||Google Inc.||Fake skip evaluation of synonym rules|
|US8910037 *||Feb 28, 2012||Dec 9, 2014||Google Inc.||Comparing text pages using image features based on word positions|
|US8938455||Sep 12, 2012||Jan 20, 2015||Facebook, Inc.||System and method for determining a homepage on the world-wide web|
|US8949256 *||Sep 12, 2012||Feb 3, 2015||Facebook, Inc.||System and method for identifying an owner of a web page on the World-Wide Web|
|US8954442||Sep 15, 2012||Feb 10, 2015||Facebook, Inc.||System and method for classifying an electronic document|
|US8959103||May 25, 2012||Feb 17, 2015||Google Inc.||Click or skip evaluation of reordering rules|
|US8965875||Apr 10, 2012||Feb 24, 2015||Google Inc.||Removing substitution rules based on user interactions|
|US8965882 *||Nov 22, 2011||Feb 24, 2015||Google Inc.||Click or skip evaluation of synonym rules|
|US9026514||Oct 13, 2006||May 5, 2015||International Business Machines Corporation||Method, apparatus and article for assigning a similarity measure to names|
|US9043331||Sep 15, 2012||May 26, 2015||Facebook, Inc.||System and method for indexing documents on the world-wide web|
|US9064273||Jan 7, 2013||Jun 23, 2015||Jpmorgan Chase Bank, N.A.||System and method for management and delivery of content and rules|
|US9075881||Sep 12, 2012||Jul 7, 2015||Facebook, Inc.||System and method for identifying the owner of a document on the world-wide web|
|US9141672||Dec 27, 2012||Sep 22, 2015||Google Inc.||Click or skip evaluation of query term optionalization rule|
|US9146966||Jan 7, 2013||Sep 29, 2015||Google Inc.||Click or skip evaluation of proximity rules|
|US9152698||Jan 3, 2012||Oct 6, 2015||Google Inc.||Substitute term identification based on over-represented terms identification|
|US9183300||Sep 15, 2012||Nov 10, 2015||Facebook, Inc.||System and method for geographically classifying business on the world-wide web|
|US9235624 *||Nov 9, 2012||Jan 12, 2016||Nec Corporation||Document similarity evaluation system, document similarity evaluation method, and computer program|
|US9240012||Jun 24, 2014||Jan 19, 2016||Jpmorgan Chase Bank, N.A.||Systems and methods for multifactor authentication|
|US9298833 *||Apr 2, 2015||Mar 29, 2016||Garry Carl Kaufmann||Unfailing self-correcting modified Boolean search engine|
|US9323827||Jan 30, 2008||Apr 26, 2016||Google Inc.||Identifying key terms related to similar passages|
|US9374366||Oct 10, 2013||Jun 21, 2016||Jpmorgan Chase Bank, N.A.||System and method for anti-phishing authentication|
|US9419957||Mar 17, 2014||Aug 16, 2016||Jpmorgan Chase Bank, N.A.||Confidence-based authentication|
|US20020103799 *||Dec 5, 2001||Aug 1, 2002||Science Applications International Corp.||Method for document comparison and selection|
|US20020184153 *||Jun 5, 2001||Dec 5, 2002||De Vries Jean Pierre||System and method for sharing matched interests without disclosing non-shared interests|
|US20020188601 *||Mar 27, 2001||Dec 12, 2002||International Business Machines Corporation||Apparatus and method for determining clustering factor in a database using block level sampling|
|US20030026459 *||Nov 29, 2001||Feb 6, 2003||Won Jeong Wook||System for drawing patent map using technical field word and method therefor|
|US20030028566 *||Jul 12, 2002||Feb 6, 2003||Matsushita Electric Industrial Co., Ltd.||Text comparison apparatus|
|US20030033301 *||Jun 26, 2001||Feb 13, 2003||Tony Cheng||Method and apparatus for providing personalized relevant information|
|US20030145002 *||Nov 26, 2002||Jul 31, 2003||Tzunami, Inc.||Method and system for organizing objects according to information categories|
|US20030145014 *||Jul 6, 2001||Jul 31, 2003||Eric Minch||Method and apparatus for ordering electronic data|
|US20040019846 *||Jul 24, 2002||Jan 29, 2004||Xerox Corporation||System and method for managing document retention of shared documents|
|US20040167921 *||Dec 17, 2003||Aug 26, 2004||Verdasys, Inc.||Identifying history of modification within large collections of unstructured data|
|US20040172394 *||Dec 17, 2003||Sep 2, 2004||Verdasys, Inc.||Identifying similarities within large collections of unstructured data|
|US20040177088 *||Mar 12, 2004||Sep 9, 2004||H5 Technologies, Inc., A California Corporation||Wide-spectrum information search engine|
|US20040186846 *||Jan 30, 2004||Sep 23, 2004||Birdwell John D.||Method of partitioning data records|
|US20040186920 *||Jan 30, 2004||Sep 23, 2004||Birdwell John D.||Parallel data processing architecture|
|US20040230441 *||Sep 25, 2003||Nov 18, 2004||Thomas Reilly||System and method for customizing a portal environment|
|US20050010863 *||Mar 17, 2003||Jan 13, 2005||Uri Zernik||Device system and method for determining document similarities and differences|
|US20050055342 *||Jan 21, 2004||Mar 10, 2005||Bharat Krishna Asur||Method for estimating coverage of Web search engines|
|US20050071365 *||Feb 24, 2004||Mar 31, 2005||Jiang-Liang Hou||Method for keyword correlation analysis|
|US20050081139 *||Oct 8, 2004||Apr 14, 2005||George Witwer||Clustering based personalized web experience|
|US20050132197 *||May 13, 2004||Jun 16, 2005||Art Medlar||Method and apparatus for a character-based comparison of documents|
|US20060161531 *||Apr 25, 2005||Jul 20, 2006||Fatlens, Inc.||Method and system for information extraction|
|US20060167930 *||Jan 13, 2006||Jul 27, 2006||George Witwer||Self-organized concept search and data storage method|
|US20060224584 *||Dec 27, 2005||Oct 5, 2006||Content Analyst Company, Llc||Automatic linear text segmentation|
|US20060242140 *||May 11, 2006||Oct 26, 2006||Content Analyst Company, Llc||Latent semantic clustering|
|US20060242190 *||May 11, 2006||Oct 26, 2006||Content Analyst Comapny, Llc||Latent semantic taxonomy generation|
|US20060244768 *||Dec 1, 2005||Nov 2, 2006||Humanizing Technologies, Inc.||Enhanced personalized portal page|
|US20060248570 *||Dec 1, 2005||Nov 2, 2006||Humanizing Technologies, Inc.||Customized media presentation|
|US20070156665 *||Jul 6, 2004||Jul 5, 2007||Janusz Wnek||Taxonomy discovery|
|US20070208998 *||Mar 6, 2006||Sep 6, 2007||Microsoft Corporation||Displaying text intraline diffing output|
|US20070236431 *||Mar 7, 2007||Oct 11, 2007||Sony Corporation||Light-emitting display device, electronic apparatus, burn-in correction device, and program|
|US20080034282 *||Aug 16, 2007||Feb 7, 2008||Opensource, Inc.||Device, system and method for determining document similarities and differences|
|US20080091670 *||Aug 28, 2007||Apr 17, 2008||Collarity, Inc.||Search phrase refinement by search term replacement|
|US20080091674 *||Oct 13, 2006||Apr 17, 2008||Thomas Bradley Allen||Method, apparatus and article for assigning a similarity measure to names|
|US20080109425 *||Nov 2, 2006||May 8, 2008||Microsoft Corporation||Document summarization by maximizing informative content words|
|US20080109461 *||Jan 2, 2008||May 8, 2008||University Of Tennessee Research Foundation||Parallel data processing architecture|
|US20080134195 *||Jan 2, 2008||Jun 5, 2008||University Of Tennessee Research Foundation||Parallel data processing architecture|
|US20080140643 *||Oct 11, 2007||Jun 12, 2008||Collarity, Inc.||Negative associations for search results ranking and refinement|
|US20080172402 *||Aug 10, 2007||Jul 17, 2008||University Of Tennessee Research Foundation||Method of indexed storage and retrieval of multidimensional information|
|US20080235218 *||Apr 25, 2008||Sep 25, 2008||Joel Jeffrey||Wide-spectrum information search engine|
|US20080276162 *||Nov 16, 2006||Nov 6, 2008||The University Of Tennessee Research Foundation||Method of Organizing and Presenting Data in a Table|
|US20080288428 *||Nov 16, 2006||Nov 20, 2008||The University Of Tennessee Research Foundation||Method of Interaction With an Automated System|
|US20090012789 *||Oct 17, 2007||Jan 8, 2009||Teresa Ruth Gaudet||Method and process for performing category-based analysis, evaluation, and prescriptive practice creation upon stenographically written and voice-written text files|
|US20090055361 *||Aug 27, 2008||Feb 26, 2009||Birdwell John D||Parallel Data Processing System|
|US20090055389 *||Jun 5, 2008||Feb 26, 2009||Google Inc.||Ranking similar passages|
|US20090055394 *||Jan 30, 2008||Feb 26, 2009||Google Inc.||Identifying key terms related to similar passages|
|US20090083236 *||Dec 1, 2008||Mar 26, 2009||Gary Stephen Shuster||Method and apparatus for photograph finding|
|US20090150318 *||Nov 16, 2006||Jun 11, 2009||The University Of Tennessee Research Foundation||Method of Enhancing Expert System Decision Making|
|US20090158254 *||May 13, 2008||Jun 18, 2009||Chun-Fang Cheng||Automatic creative proposal generating and filtering system and manufacturing method thereof and multiple components combining method|
|US20090182755 *||Jan 10, 2008||Jul 16, 2009||International Business Machines Corporation||Method and system for discovery and modification of data cluster and synonyms|
|US20090228296 *||Mar 4, 2009||Sep 10, 2009||Collarity, Inc.||Optimization of social distribution networks|
|US20090248753 *||Jan 3, 2008||Oct 1, 2009||Microsoft Corporation||Database management system risk assessment|
|US20100049770 *||Jun 25, 2009||Feb 25, 2010||Collarity, Inc.||Interactions among online digital identities|
|US20100114809 *||Nov 13, 2009||May 6, 2010||University Of Tennessee Research Foundation||Method of Organizing and Presenting Data in a Table|
|US20100329519 *||Sep 2, 2010||Dec 30, 2010||Gary Stephen Shuster||Method and apparatus for photograph finding|
|US20110035403 *||Jun 14, 2010||Feb 10, 2011||Emil Ismalon||Generation of refinement terms for search queries|
|US20110173197 *||Jan 12, 2010||Jul 14, 2011||Yahoo! Inc.||Methods and apparatuses for clustering electronic documents based on structural features and static content features|
|US20110191369 *||Mar 8, 2011||Aug 4, 2011||Lexisnexis, A Division Of Reed Elsevier Inc.||Systems and Methods for Providing A Validation Tool|
|US20110219000 *||Nov 6, 2009||Sep 8, 2011||Yukitaka Kusumura||Search apparatus, search method, and recording medium storing program|
|US20120023102 *||Aug 30, 2011||Jan 26, 2012||Veveo, Inc.||Methods and systems for dynamically rearranging search results into hierarchically organized concept clusters|
|US20120254209 *||Mar 23, 2012||Oct 4, 2012||Casio Computer Co., Ltd.||Searching method, searching device and recording medium recording a computer program|
|US20130144870 *||Dec 2, 2011||Jun 6, 2013||Anjani Gupta||Results returned for list-seeking queries|
|US20130173626 *||Sep 12, 2012||Jul 4, 2013||Facebook, Inc.||System and method for geographically organizing and classifying businesses on the world-wide web|
|US20130191367 *||Feb 20, 2013||Jul 25, 2013||Facebook, Inc.||System and method for geographically organizing and classifying businesses on the world-wide web|
|US20130191410 *||Nov 9, 2012||Jul 25, 2013||Nec Corporation||Document similarity evaluation system, document similarity evaluation method, and computer program|
|US20140095519 *||Mar 1, 2013||Apr 3, 2014||Oracle International Corporation||Evaluating XML Full Text Search|
|US20140350941 *||May 21, 2013||Nov 27, 2014||Microsoft Corporation||Method For Finding Elements In A Webpage Suitable For Use In A Voice User Interface (Disambiguation)|
|US20150073944 *||Nov 17, 2014||Mar 12, 2015||Craze, Inc.||Method and system for classification of venue by analyzing data from venue website|
|CN102737113A *||Mar 29, 2012||Oct 17, 2012||卡西欧计算机株式会社||Searching method and searching device|
|EP1170674A2 *||Nov 21, 2000||Jan 9, 2002||LION Bioscience AG||Method and apparatus for ordering electronic data|
|EP1170674A3 *||Nov 21, 2000||Apr 17, 2002||LION Bioscience AG||Method and apparatus for ordering electronic data|
|EP2595065A1||Nov 15, 2011||May 22, 2013||Kairos Future Group AB||Categorizing data sets|
|WO2001011487A2 *||Aug 4, 2000||Feb 15, 2001||Board Of Trustees Of The University Of Illinois||Apparatus, method and product for multi-attribute drug comparison|
|WO2001011487A3 *||Aug 4, 2000||Dec 24, 2003||Univ Illinois||Apparatus, method and product for multi-attribute drug comparison|
|WO2002005084A3 *||Jul 6, 2001||Apr 25, 2002||Lion Bioscience Ag||Method and apparatus for ordering electronic data|
|WO2004066086A3 *||Jan 21, 2004||Jan 20, 2005||Verdasys Inc||Identifying similarities and history of modification within large collections of unstructured data|
|WO2013072258A1||Nov 9, 2012||May 23, 2013||Kairos Future Group Ab||Unsupervised detection and categorization of word clusters in text data|
|WO2013124520A1 *||Feb 22, 2012||Aug 29, 2013||Nokia Corporation||Adaptive system|
|U.S. Classification||707/737, 707/E17.091, 709/217, 715/234, 707/773, 707/823, 707/999.005, 707/749, 707/750, 707/999.006|
|Cooperative Classification||Y10S707/99935, Y10S707/99936, G06F17/3071|
|Oct 8, 1998||AS||Assignment|
Owner name: MANTRA TECHNOLOGIES, INC., CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:HILSENRATH, OLIVER A.;CARMEL, RON;ARIEL, HAGAI;REEL/FRAME:009408/0956;SIGNING DATES FROM 19971006 TO 19971009
|Feb 5, 2003||REMI||Maintenance fee reminder mailed|
|Jul 21, 2003||LAPS||Lapse for failure to pay maintenance fees|
|Sep 16, 2003||FP||Expired due to failure to pay maintenance fee|
Effective date: 20030720